On Thu, 17 Jun 2021 21:40:07 -0400
Vivek Goyal <vgo...@redhat.com> wrote:
> On Thu, Jun 17, 2021 at 04:15:18PM +0200, Greg Kurz wrote:
> > A well behaved FUSE client uses FUSE_CREATE to create files. It isn't
> > supposed to pass O_CREAT along a FUSE_OPEN request, as documented in
> > the "fuse_lowlevel.h" header :
> >
> > /**
> > * Open a file
> > *
> > * Open flags are available in fi->flags. The following rules
> > * apply.
> > *
> > * - Creation (O_CREAT, O_EXCL, O_NOCTTY) flags will be
> > * filtered out / handled by the kernel.
> >
> > But if it does anyway, virtiofsd crashes with:
> >
> > *** invalid openat64 call: O_CREAT or O_TMPFILE without mode ***: terminated
>
This is also the consequence of virtiofsd being compiled with
-D_FORTIFY_SOURCE=2. Without that, no abort but arbitrary data
is passed as mode_t argument to the openat() syscall instead.
> So did you hit this error with current fuse client. If yes, that means
> client needs fixing as well?
>
I've patched the client to cause this:
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -28,6 +28,7 @@ static int fuse_send_open(struct fuse_mount *fm, u64 nodeid,
memset(&inarg, 0, sizeof(inarg));
inarg.flags = open_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
+ if (opcode == FUSE_OPEN) inarg.flags |= O_TMPFILE;
if (!fm->fc->atomic_o_trunc)
inarg.flags &= ~O_TRUNC;
> Or you are doing this fix based on comment in fuse_lowlevel.h.
>
> I am wondering why this protocl restriction is there that open()
> path should not be able to honor O_CREAT.
>
It isn't a protocol restriction IMHO. The distinction between file
creation and file opening has always been there since the start.
Older versions of the protocol would send FUSE_MKNOD to create a
file and then send FUSE_OPEN to open it. Because this was racy,
FUSE_CREATE was introduced at some point to do both operations
atomically.
Question is : what would be the semantics of O_CREAT in FUSE_OPEN ?
> Vivek
>
> >
> > This is because virtiofsd ends up passing this flag to openat() without
> > passing a mode_t 4th argument which is mandatory with O_CREAT, and glibc
> > aborts.
> >
> > The offending path is:
> >
> > lo_open()
> > lo_do_open()
> > lo_inode_open()
> >
> > Other callers of lo_inode_open() only pass O_RDWR and lo_create()
> > passes a valid fd to lo_do_open() which thus doesn't even call
> > lo_inode_open() in this case.
> >
> > Specifying O_CREAT with FUSE_OPEN is a protocol violation. Check this
> > in lo_open() and return an error to the client : EINVAL since this is
> > already what glibc returns with other illegal flag combinations.
> >
> > The FUSE filesystem doesn't currently support O_TMPFILE, but the very
> > same would happen if O_TMPFILE was passed in a FUSE_OPEN request. Check
> > that as well.
> >
> > Signed-off-by: Greg Kurz <gr...@kaod.org>
> > ---
> > tools/virtiofsd/passthrough_ll.c | 6 ++++++
> > 1 file changed, 6 insertions(+)
> >
> > diff --git a/tools/virtiofsd/passthrough_ll.c
> > b/tools/virtiofsd/passthrough_ll.c
> > index 49c21fd85570..14f62133131c 100644
> > --- a/tools/virtiofsd/passthrough_ll.c
> > +++ b/tools/virtiofsd/passthrough_ll.c
> > @@ -2145,6 +2145,12 @@ static void lo_open(fuse_req_t req, fuse_ino_t ino,
> > struct fuse_file_info *fi)
> > return;
> > }
> >
> > + /* File creation is handled by lo_create() */
> > + if (fi->flags & (O_CREAT | O_TMPFILE)) {
> > + fuse_reply_err(req, EINVAL);
> > + return;
> > + }
> > +
> > err = lo_do_open(lo, inode, -1, fi);
> > lo_inode_put(lo, &inode);
> > if (err) {
> > --
> > 2.31.1
> >
>
